Assessment and Testing
What this domain tests
Psychometric properties, score conversion, test selection, and the ethical use of instruments, often with a small computation.
Core rules
- Reliability types: test-retest (stability), alternate forms, internal consistency (split-half, KR-20, Cronbach's alpha), inter-rater. Standard error of measurement: SEM = SD x sqrt(1 - r). Build confidence intervals as score +/- (z)(SEM).
- Validity: face, content, construct (convergent and discriminant), criterion (concurrent and predictive). Reliability is necessary but not sufficient for validity.
- Standard scores: z (M=0, SD=1), T (M=50, SD=10), IQ and most standard scores (M=100, SD=15), scaled scores (M=10, SD=3), stanines (M=5, SD=2, range 1-9), percentile rank.
- Familiar instruments: MMPI-3, MCMI-IV, BDI-II, WISC-V (verbal comprehension, visual spatial, fluid reasoning, working memory, processing speed), WAIS-IV (verbal comprehension, perceptual reasoning, working memory, processing speed), Vineland, BASC-3, Strong Interest Inventory.
- ACA Section E: only use tests you are competent with, obtain informed consent, consider cultural fairness and norm-group match, and maintain test security.
Worked example
A client scores 1.5 SD above the mean on a test with M=100, SD=15. Reasoning: z = 1.5, so the standard score is 100 + 1.5(15) = 122.5; T = 50 + 1.5(10) = 65; percentile is approximately 93. A score 1 SD below the mean would be 85, T = 40, percentile about 16.
Traps that cost marks
- A percentile is not percent correct; the 50th percentile equals the mean.
- A reliable test can still be invalid for the intended purpose.
- Grade equivalents overstate precision.
- Screening instruments produce false positives and negatives; check base rates.